MSR Decode | Kalshi NFL protection | October 24, 2026

These are normalized extracts from retained public API responses, not customer account statements. Source raw_sha256/sha256 fields identify the original response bytes; file-hashes.json hashes the published extracts. Local storage paths and HTTP headers are omitted. Public request URLs, market identifiers and recorded selected sides are retained. Original batch responses are retained privately.

Exact pairs match normalized selected-leg signatures and YES/NO sides, using official protected-to-ordinary product mappings and verified constituent twins. 55 pairs were recovered. The 42 five-second global trade-feed windows total 210 seconds, at 60, 15 and 5 minutes before 14 kickoff anchors for 29 scheduled NFL games October 11-22. Nine windows reuse earlier captures; 33 were newly collected. The global feed includes other games/sports. These are overlapping exploratory frames, not independent population samples.

For each protected print, compare the latest strictly prior ordinary print on the same trading side inside the same window. Retain every tie at that ordinary timestamp. Median differences within each eligible pair, then take the median across pairs (equal pair weight). 46 eligible protected execution groups produce 59 differences across 28 pairs; 27 pairs have none. All eligible protected prints are YES-side. All 59 differences are positive. Median of pair medians: $0.0024 = 0.24 cents per contract. Combining timestamp ties first produces the same result. Median time separation is 1.0006815 seconds; maximum 3.92053 seconds. Quantities differ, and these are not simultaneous quotes, fees or a verified premium paid by a particular customer.

Settlement values are per $1 face value. For NO-selected legs use 1 minus the YES settlement. Only finalized markets with recorded values count as resolved. Of 55 exact combo pairs, 48 are zero/zero, six one/one, and one 0.31/zero. Of 484 protected-containing combos, 427 pay zero, 43 one, 11 fractionally and three are unresolved. Fifty have a fractional protected selected-side constituent: 39 pay zero with another observed losing selection; 11 pay fractionally. This is not an injury count, customer rescue rate or expected-value calculation. The Skattebo example overlaps this research.

The separate October 22 New England-Chicago catalog has 349 exact individual-market pairs (237 zero/zero,112 one/one), plus 50 ordinary-only markets. These are not 349 combo pairs. The selected game was known in earlier research; it is not a blind experiment. Source ledger preserves all catalog requests.

Captured market terms are later API records, not contemporaneous order screens. Public participation reporting supports the Skattebo first-half-only reconstruction, not a complete snap census or proof of deliberate benching. Legacy contract-version assignment and customer quote/fill/fee/credit histories remain unresolved. No overpricing, deception, wrongful settlement or customer harm is established.